Does decoherence explain all instances of wave function collapse? Specifically, how can decoherence explain the appearance of flecks of metallic silver on a photographic plate when exposed to the very weak light of a distant star?
EDIT: Perhaps the advocates of decoherence need some context for this question. There is a certain definite quantity of energy on the order of one or two eV to drive the chemical reaction
2AgBr -> 2Ag +  Br2
This is the reaction responsible for the fleck of silver on the photographic plate. The amount of energy is far greater than can be accounted for in any realistic time frame by the classical e-m wave energy of the light of a distant star. 
Any explanation must explain where this energy comes from. How does "decoherence" claim to do this? I have heard over and over again that there is a matrix which is diagonalized, but no one has so much as volunteered to say just what matrix they are talking about. Is it, for example, the matrix of position states of the photon? Or perhaps it is the oxidation states of the silver atom? And I would really like a better explanation of how the matrix is "diagonalized" than to simply repeat that it is in "thermal contact with the environment."  
EDIT: I have reviewed the comments again and I find that no one has come close to dealing with the question. I cannot find anything wrong in the way I have asked it so far, so I am posting this edit as my only means to prompt people to attempt an answer.
 A: Almost phenomenologically, "the appearance of flecks of metallic silver on a photographic plate" is a thermodynamic transition that happens at different rates depending on the details of how the plate is prepared and on details of the exposure of the photographic plate. Such thermodynamic transitions are often correlated in nontrivial ways. All QM has to do to be Useful is to model or describe the statistics of the thermodynamic transitions. [Note that my introduction of the idea of a thermodynamic transition makes my claim here "theory-laden", not quite phenomenological, at least to that extent.]
Explanation is not necessary for Usefulness. One topic of research in Philosophy of Physics has been to try to determine what makes a model "explanatory", which IMO has been rather inconclusive. Models may be more or less Useful for many different reasons, including tractability and directness of reference between elements of the theory and elements of experimental signal data. Note that a class of models may seem explanatory for 50 years even if it is the phlogiston theory, if the model is superficially nice in whatever ways.
Which brings me to my Answer, which I'm pretty sure you won't find Useful, which is that Decoherence doesn't explain particularly well, whatever that means, partly because it's not a very tractable approach. Decoherence seems to have fairly direct referents, which perhaps is what makes it appeal to some people quite strongly. The same is true of "wave function collapse": it's possible to structure experimental data taking wave function collapse as a fundamental modeling strategy, but so far no-one has produced a mathematization that is enough more Useful than just dealing with the statistics of thermodynamic events. There are people who think it illuminates what we're doing with QM in ways that might lead to a better mathematical formulation of the whole theory, but, I think, nothing yet.
In a similar vein, you may notice that Particle Physics is more often called High Energy Physics than it used to be, which seems to me to reflect the realization, not uniformly acknowledged, that the explanation of tracks of obviously related thermodynamic events in detectors as "caused by particles" is weakened by the many low-energy experiments that show that the concept of a particle cannot be that simple. As of now, Quantum fields are as likely to be the locus of descriptions of experiments.
I'm curious whether you can knock down this argument, such as it is. I think you're looking at this all wrong, but of course it may be me. That I've worked on this for a long time doesn't guarantee much.
EDIT (a long comment, in response to Marty's comment that first mentions "Quantum Siphoning"): I take the Wave Function and operators to be a good way to generate probability measures. The empirical success comes from the probabilities being able to be good models for (or descriptions of) statistics of raw experimental data. I take it that probabilities do not cause individual events, they describe sets of events (propensity interpretations of probability notwithstanding). [If we go the Wigner function route --which I don't, except as a mathematical equivalence, because I think it obscures the relationship to empirical data-- the wave function is just a generalized probability function that sometimes has negative values.] If one wants to change probability distributions as a result of experience, instead of taking other approaches to statistics, then one should use something like Bayes' rule, which in general doesn't just change the probability from 0.615802 to 0 or to 1. "collapse" of the wave function adds an extra level of structure to the concept of a probability distribution that I think just doesn't fit well, as Mathematics. If people want to use "collapse", I think it has to be done somehow differently. It's possible that a propensity interpretation could work for you, but I think we would then quite quickly get far enough apart that we can't talk to each other.
I think I prefer my description of individual events (and we may just have to accept that this is a sticking point)-- that we should say that the individual events are "thermodynamic transitions", whatever that means, leaving a causal account of how that happens for the future. The concept of thermodynamic transitions is the historical concept from Physics that I think fits the case. A thermodynamic event implicitly invokes at least a large number, perhaps an infinite number of degrees of freedom, to explain what happens when there is an apparent discontinuity, it introduces a degree of complexity that is hard to manage mathematically, which definitely has its problems. Decoherence also introduces an infinite number of degrees of freedom, but I think by introducing the environment in the way it does it doesn't adequately embrace the complexity of the photographic plate. I think your description of what happens in a photographic plate accepts that complexity, but then looks to make "collapse" of a quantum state, which has nowhere near as much structure as the photographic plate, be an explanation of what is happening. It's important that it not be brushed under the table, but we can measure where and when thermodynamic events happen without knowing how they happen.
I hope that's helpful. I expect no-one else is listening!
A: The answer is simple: Decoherence doesn't explain wavefunction collapse and it couldn't possibly do such a thing. Decoherence and collapse are complementary phenomenons, but they are fundamentally different
Decoherence at most will make all interference terms to have essentially random phases, which will average out to zero, leaving only squared-positive probabilities on the diagonal. This mixed-state is usually compared with a classical probability distribution in microstates, and people usually like to make the judgement jump to say that decoherence 'produced' a classical limit. But the truth is that decoherence didn't do such a thing. The probabilities, even if they can be interpreted as a classical distribution, their uncertainties are inherently quantum. Individual measurements (in the example you used, a specific point in a silver photographic plate) will still be individual collapses. Only after having a statistically significant sample of eigenstates, you recover a classical probability distribution that will match the Density matrix (either decohered with the environment, or not)
A: 
Any explanation must explain where this energy comes from.

The energy comes from the energy of a single photon that has come all that way from a distant star. E=h*nu , nu the frequency. 

How does "decoherence" claim to do this?

Decoherence has little  to do with single photons or particles. It mainly has to do with a many particle system where the particles are coherent, i.e . are the complete solution of a quantum mechanical equation for a many body system, a single state function, where all the phases are defined by the boundary conditions of the creation of this state function.

I have heard over and over again that there is a matrix which is diagonalized, but no one has so much as volunteered to say just what matrix they are talking about. Is it, for example, the matrix of position states of the photon? 

A laser beam for example is coherent, which means that each photon represented by a vector in a matrix representation will have off diagonal elements with the other photons, well defined, i.e. the phases are fixed and well defined. When the beam decoheres the off diagonal elements tend to zero and there is only the individual photon in the diagonal.

Or perhaps it is the oxidation states of the silver atom?

The density matrix as far as the photon is concerned is irrelevant int this situation. As far as the reactions initiated with the silver atoms it may be formulated since they are a many body state, but simple chemistry is sufficient.

And I would really like a better explanation of how the matrix is "diagonalized" than to simply repeat that it is in "thermal contact with the environment." 

Well, here is a blog entry that treats the density matrix classically and quantum mechanically. I will read it myself, to nail down how the phases tend to zero exactly. Up to now it seemed a reasonable guess for a many body randomizes system.
The photons hitting the silver from a star is a single photon. Its energy, transmitted to the molecular system initiates a chemical reaction, which can be studied quantum mechanically and with a density matrix formalism, but is irrelevant to the photon's history.
Now if we assume that the photon from a de-excitation of an atom in the star light years away has not interacted on the way, and this is probable, as we do see absorption spectra from stars, then the state function atom-photon "collapses" when the photon hits the silver, i.e. when it interacts the first time. "Collapse" is a way of saying "interacts", The probability function (which is the state function squared,) tells us how the photon was created and where it comes from etc. The silver spec  gives  one hit in building up experimentally the probability distribution.
